Latest / The Official SaaStr Podcast: SaaS | Founders | Investors / SaaStr Labs: Replit v3 ... Our Latest AI SDR Crushes It ... And 300,000 AI Startup Valuations
Transcript
- 0:00SPEAKER_00: We've done over 300,000 startup valuations on our AI calculator.
- 0:05And we've also graded and helped improve almost 1,000 pitch
- 0:08decks in less than two weeks.
- 0:09SPEAKER_00: I didn't know.
- 0:10SPEAKER_00: And basically,
- 0:11SPEAKER_00: everyone said they were somewhat similar a couple months ago,
- 0:13SPEAKER_00: which I think was true a couple months ago.
- 0:15SPEAKER_00: But Replit was the one that you could go end-to-end.
- 0:18And deep down,
- 0:18actually,
- 0:19Replit didn't even use the most latest versions of Claude at
- 0:23Anthropic.
- 0:23SPEAKER_00: It uses, actually, the N-1 version
- 0:26SPEAKER_00: And that could still do a lot of things.
- 0:27SPEAKER_00: But then recently they launched V3.
- 0:30And unless you're deep in the ecosystem or nerdy,
- 0:33you might not have seen it.
- 0:35But I think it is the first glimpse of the future,
- 0:37of what is coming next.
- 0:39SPEAKER_00: But now it has agents that can manage agents.
- 0:42SPEAKER_00: So if you're on X and you're on the internet,
- 0:44SPEAKER_00: you'll hear lots of folks talking about agents,
- 0:46SPEAKER_00: managing agents.
- 0:47Now,
- 0:48pretty crazy that you have an AI autonomously managing other
- 0:51AIs in Replit.
- 0:53SPEAKER_00: If you doubted that was coming...
- 0:55SPEAKER_00: there's a hint of it that's already here.
- 0:56SPEAKER_00: It was crazy.
- 0:57SPEAKER_00: It was crazy that they could do this for up to three hours.
- 1:00SPEAKER_00: They could do it autonomously and they could debate with each other.
- 1:03But then I will say for the B2B folks that might be watching or
- 1:06listening,
- 1:08SPEAKER_00: another learning from this V3,
- 1:09SPEAKER_00: it's so cool,
- 1:09SPEAKER_00: but man,
- 1:10SPEAKER_00: it's different.
- 1:11The biggest reason folks got upset is because V3 is just
- 1:13different and we weren't asked.
- 1:15Replet didn't offer an opportunity for folks to use V2 that
- 1:18still wanted to use V2,
- 1:19SPEAKER_00: nor did chat GBT when it rolled out GBT5 and
- 1:23SPEAKER_00: It's just interesting when you're moving at AI speed, like you have to.
- 1:26SPEAKER_00: Replitz has deep competition.
- 1:28SPEAKER_00: You got to move this quickly.
- 1:30SPEAKER_00: We started with one AI outbound SGR.
- 1:33SPEAKER_00: Now we have three for three different workflows.
- 1:35SPEAKER_00: It's not the bubble that really distinguishes anything anymore.
- 1:38SPEAKER_00: They all look the same.
- 1:39But what's very clever about this one is it integrates with all
- 1:42of our data.
- 1:43SPEAKER_00: It integrates with all of our Salesforce data.
- 1:45And for better or worse,
- 1:46we use Marketo,
- 1:47which is very dated for marketed automation.
- 1:49SPEAKER_00: There's nothing more soul crushing to me than when I...
- 1:51SPEAKER_00: Go to an app.
- 1:52SPEAKER_00: I click contact me, which is
- 1:55SPEAKER_00: I hate as it is.
- 1:56SPEAKER_00: You probably do too.
- 1:57And then I finally get connected to sales and all it is some
- 2:00kid that knows nothing
- 2:01that's qualifying me to decide if it's worth their time for an
- 2:04AE to talk to me.
- 2:05SPEAKER_00: Take a pause.
- 2:06SPEAKER_00: Go to your own website.
- 2:07SPEAKER_00: Try to buy your own product, folks.
- 2:09SPEAKER_00: I bet it's worse than what I just described.
- 2:11SPEAKER_00: I bet you can't get great answers in real time.
- 2:13SPEAKER_00: Ask Digital Jason any questions about how to build a comp plan,
- 2:17SPEAKER_00: what to do about a struggling VP of marketing,
- 2:19SPEAKER_00: what to do if you're running out of money,
- 2:20SPEAKER_00: whatever it is.
- 2:21SPEAKER_00: Ask Digital Jason and the answers will be insanely great.
- 2:24SPEAKER_00: And folks want to set and forget does not work, folks.
- 2:27SPEAKER_00: Set and forget, you have to relentlessly iterate on these AI platforms.
- 2:30SPEAKER_00: And that's why ours is pretty darn good.
- 2:33SPEAKER_00: Brex kept sending us
- 2:35SPEAKER_00: an email request to please give them,
- 2:36SPEAKER_00: deposit $5 million with them to maintain our credit limits.
- 2:40SPEAKER_00: Now, the emails were crazy.
- 2:41SPEAKER_00: They were driven by, written by AI.
- 2:43SPEAKER_00: If they'd been written more clearly, it wouldn't have been so funny.
- 2:46It was these alarming emails that I got from like the chief
- 2:49credit officer,
- 2:50SPEAKER_00: the chief operating officer.
- 2:52SPEAKER_00: They kept being from different officers,
- 2:54SPEAKER_00: seemingly saying we had to immediately deposit $5 million at Brex.
- 2:58SPEAKER_00: They were so funny.
- 2:58SPEAKER_00: I didn't take it too seriously, but I just wanted to learn.
- 3:01SPEAKER_00: You have to train them relentlessly.
- 3:03SPEAKER_00: And even more,
- 3:04SPEAKER_00: it's not what a lot of people think,
- 3:05SPEAKER_00: which is massive amounts of data upfront,
- 3:07SPEAKER_00: and then you could ignore it.
- 3:08SPEAKER_00: No, it's a body of content, which you have enough of to train your agent.
- 3:11SPEAKER_00: And then every day,
- 3:13SPEAKER_00: relentlessly watching it,
- 3:14SPEAKER_00: seeing it,
- 3:14SPEAKER_00: QAing it,
- 3:15SPEAKER_00: testing it,
- 3:16and making it better every single day so that whatever problems
- 3:18you have don't
- 3:19SPEAKER_00: recur.
- 3:21SPEAKER_00: Training AI is as much or more work than training a human.
- 3:24SPEAKER_00: But if you do it reasonably correct,
- 3:26SPEAKER_00: they won't keep making the same mistakes again and again,
- 3:28SPEAKER_00: which is so wonderful.
- 3:30SPEAKER_00: But check it out at saster.ai.
- 3:31We've got about 10 or 20% of our traffic there,
- 3:34but about 30,000 people a month are using it.
- 3:37SPEAKER_00: And probably the most popular stuff are our tools.
- 3:40SPEAKER_00: There are tools.
- 3:41SPEAKER_00: We launched it.
- 3:42SPEAKER_00: We've done over 300,000 free valuations on it.
- 3:46SPEAKER_00: It is so cool.
- 3:47SPEAKER_00: So try it for free.
- 3:48SPEAKER_00: Again,
- 3:48SPEAKER_00: if you're going 300 to 3,000%,
- 3:51SPEAKER_00: try it and we'll tell you what your startup's probably worth.
- 3:54SPEAKER_00: How do you review 1,000 VC pitch decks?
- 3:56SPEAKER_00: Look, a long topic, but what I learned is offload it, offload it.